New Research Highlights Untapped Potential for Satellite and Machine Learning Systems in Sugarcane Health Monitoring

A recent study explored an underused approach to monitor the health of sugarcane. This study, published in the journal, Computers and Electronics in Agriculture, demonstrated how satellite-based spectroscopy combined with machine learning (ML) can improve large-scale monitoring of sugarcane (1). Authored by Ethan Kane Waters, Carla Chia-Ming Chen, and Mostafa Rahimi Azghadi, all of James Cook University in Australia, this study shows how spectroscopy can improve monitoring of sugarcane (1).

Both Australia and the United States are major sugar producers. For the United States, annual production averages 9.0 million short tons, raw value (STRV) since 2016/17 (2). Production growth has been driven by technological advancements, improved crop varieties, and expanded acreage, though yields still fluctuate with weather conditions (2). Florida and Louisiana are the primary sugarcane-producing states, with Florida benefiting from favorable growing conditions and Louisiana expanding due to varietal and technological improvements (2,3). Texas contributes modestly, while Hawaii ceased production in 2016 (2).

Sugar cane stalks with sugar cane plantation background. | Image Credit: © Paitoon - stock.adobe.com

Meanwhile, Australia farms most of its sugarcane in Queensland (approximately 95%) (4). Australia produces approximately 4 million tons every year (4). Sugarcane is an intuitive crop to grow because it requires immediate processing after harvest due to rapid sucrose deterioration. As a result, sugarcane growers require methods that allow them to conduct efficient harvesting and processing operations.

In their study, the researchers discuss the use of satellite imagery in crop monitoring and how it can benefit sugarcane operations. The authors began their article by discussing how remote sensing the machine learning have been increasingly used in agriculture (1). However, there is a lack of integrated systems capable of detecting sugarcane diseases and pests at a large scale (1).

To demonstrate this point, the researchers spotlight decades of previously published papers, noting several areas where current sugarcane monitoring systems fall short. These include the inconsistent use of vegetation indices, failure to account for environmental and crop-specific factors that affect spectral reflectance, and the limited use of ML for multi-disease classification (1). For instance, crucial variables such as crop age, soil type, water content, sugarcane variety, and recent weather patterns have been largely overlooked in existing ML models, even though they directly impact reflectance readings and, subsequently, detection accuracy (1).

As a result, the researchers argue that an overreliance on raw spectral data has led researchers not to use multi-index models, which can be more accurate. The authors recommend that future research explore combinations of indices, such as those detecting moisture content and chlorophyll levels, to better identify early signs of stress, disease, or infestation (1).

Another limitation in this space is that not many researchers have compared satellite-based spectroscopy to drone-based spectroscopy. Although drones offer superior spatial resolution, they are less scalable than satellites (1). The researchers advocate for direct comparisons of both platforms, emphasizing that a clear understanding of the trade-offs in cost, accuracy, and deployment can help us to understand what approach would be more suitable for crop monitoring applications (1).

Another pressing issue is the need for disease detection models that incorporate time-series data to identify threats at different sugarcane growth stages. The review highlights a gap in the literature surrounding how the crop’s life cycle influences disease detectability. The research team recommends that a greater focus needs to be on temporal disease modeling (1).

The authors conclude their article by discussing how the development of user-friendly software is key to advancing crop monitoring applications. Ideally, scientists would be able to build the software so it can automatically acquire satellite imagery, calculate relevant indices, and present disease alerts through an intuitive dashboard for farmers and agronomists (1). To encourage industry adoption, a cost-benefit analysis comparing current manual methods with proposed satellite-driven solutions is also deemed essential (1).

Overall, this study highlights that to improve sugarcane monitoring, precision agriculture models need to be improved. Developing these tools, the authors argue, would help sugarcane producers reduce crop loss and improve food security in sugar-producing regions in Australia, the United States, and around the globe (1).

References

1. Waters, E. K.; Chen, C. C.-M.; Azghadi, M. R. Sugarcane Health Monitoring with Satellite Spectroscopy and Machine Learning: A Review. Com. Elect. Agric. 2025, 229, 109686. DOI: 10.1016/j.compag.2024.109686
2. U.S. Department of Agriculture, Sugar and Sweeteners – Background. USDA.gov. Available at: https://www.ers.usda.gov/topics/crops/sugar-and-sweeteners/background (accessed 2025-05-06).
3. Wetzel, W. Near-Infrared Spectroscopy Enhances Sugar Cane Quality Analysis, Study Finds. Spectroscopy. Available at: https://www.spectroscopyonline.com/view/near-infrared-spectroscopy-enhances-sugar-cane-quality-analysis-study-finds (accessed 2025-05-06).
4. Sugar Nutrition Resource Centre, Growing Sugarcane. Sugar Nutrition Resource. Available at: https://www.sugarnutritionresource.org/news-articles/growing-sugar-cane#:~:text=In%20Australia%2C%20sugar%20cane%20is%20farmed%20predominately,of%20sugar%20and%20employed%20over%2022%2C000%20people. (accessed 2025-05-06).